← Home
This Week in Startups · July 15, 2026 · 01:14:04

A Startup Is Trying to Buy PayPal… Craziest Deal of 2026! | E2312

In the wake of news that Stripe has made a $53 billion offer for PayPal, Alex Wilhelm sits down with investors Eric Bahn and Jeff Morris Jr. for a venture capital roundtable. They dissect what the deal means for fintech and why it underscores the growing trend of startups avoiding public markets. The conversation covers when founders should go all-in ('burn the boats'), the potential of physical AI, the role of AI within venture capital itself, and the rise of company-specific AI evaluations. The group brings both founder and investor perspectives to these timely topics.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Highlights

Stripe's Attempt to Acquire PayPal00:01:35
Eric Bahn and Jeff Morris Jr. discuss the news that Stripe has made a $53 billion offer to acquire PayPal, a deal that would dramatically reshape the fintech landscape.

Editorial

Founders Avoiding IPOs00:06:29
The discussants unpack why founders increasingly prefer to stay private rather than go public, citing factors like regulatory burden, public market short-termism, and the availability of private capital.
Burning the Boats00:15:16
Using the metaphor of 'burning the boats,' the group discusses the moments when startups should commit fully to a single path, as illustrated by the story of Fin selling to Salesforce.
VC Needs an AI Revolution00:21:16
The panelists discuss whether venture capital firms themselves need to undergo an AI transformation to remain competitive, or if the current hype is overblown.
Physical AI as the Next Frontier
The group identifies 'physical AI' — AI applied to real-world systems like robotics, manufacturing, and hardware — as a fertile area for new startup investment.
Company-Specific AI Evals
The roundtable notes a new practice: startups building custom evaluation frameworks for their own AI models and use cases, rather than relying on generic benchmarks.
Was this useful?